Rethinking the hippocampal cognitive map as a meta-learning computational module

Ambrogioni, Luca; Ólafsdóttir, Hildur

doi:10.1016/j.tics.2023.05.011

Cited by 8 publications

(3 citation statements)

References 66 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Probabilistic recall can be used to encode Bayesian uncertainty, with several possible memories being simultaneously co-activated by parallel denoising processes. This form of encoding can be used to solve information gathering, exploration and other meta-learning problems [43] and has been suggested to be central to the functioning of the mammalian hippocampus [44].…”

Section: Probabilistic Recallmentioning

confidence: 99%

In Search of Dispersed Memories: Generative Diffusion Models Are Associative Memory Networks

Ambrogioni

2024

Entropy

Self Cite

View full text Add to dashboard Cite

Uncovering the mechanisms behind long-term memory is one of the most fascinating open problems in neuroscience and artificial intelligence. Artificial associative memory networks have been used to formalize important aspects of biological memory. Generative diffusion models are a type of generative machine learning techniques that have shown great performance in many tasks. Similar to associative memory systems, these networks define a dynamical system that converges to a set of target states. In this work, we show that generative diffusion models can be interpreted as energy-based models and that, when trained on discrete patterns, their energy function is (asymptotically) identical to that of modern Hopfield networks. This equivalence allows us to interpret the supervised training of diffusion models as a synaptic learning process that encodes the associative dynamics of a modern Hopfield network in the weight structure of a deep neural network. Leveraging this connection, we formulate a generalized framework for understanding the formation of long-term memory, where creative generation and memory recall can be seen as parts of a unified continuum.

show abstract

Section: Probabilistic Recallmentioning

confidence: 99%

In Search of Dispersed Memories: Generative Diffusion Models Are Associative Memory Networks

Ambrogioni

2024

Entropy

Self Cite

View full text Add to dashboard Cite

show abstract

“…The presentation of spatial cognitive deficits following vestibular dysfunction naturally implicates the hippocampus (HPC), since it is known to be important for spatial cognition (see Ambrogioni and Ólafsdóttir, 2023 , for a recent review). Indeed, through the vestibular nucleus and presumably also the cerebellum, vestibular information is transmitted to the HPC (see Hitier et al, 2014 for a review; see Hitier et al, 2021 for a recent example).…”

Section: Introductionmentioning

confidence: 99%

“…Although some of this evidence has not controlled for the potential effects of concurrent hearing loss, many recent studies have, and it has become obvious that both hearing loss and vestibular loss contribute to cognitive dysfunction, with loss of function in both sensory systems exacerbating the situation (see Smith, 2022a,b for a recent discussion). The presentation of spatial cognitive deficits following vestibular dysfunction naturally implicates the hippocampus (HPC), since it is known to be important for spatial cognition (see Ambrogioni and Ólafsdóttir, 2023, for a recent review). Indeed, through the vestibular nucleus and presumably also the cerebellum, vestibular information is transmitted to the HPC (see Hitier et al, 2014 for a review; see Hitier et al, 2021 for a recent example).…”

Section: Introductionmentioning

confidence: 99%

Interpreting the meaning of changes in hippocampal volume associated with vestibular loss

Smith

2023

Front. Integr. Neurosci.

View full text Add to dashboard Cite

Many studies have documented cognitive deficits, especially spatial cognitive deficits, in patients with some form of vestibular loss. Almost 20 years ago, hippocampal (HPC) atrophy was reported to be correlated with spatial memory deficits in such patients and the idea has gradually emerged that HPC atrophy may be causally responsible for the cognitive deficits. However, the results of studies of HPC volume following vestibular loss have not always been consistent, and a number of studies have reported no evidence of HPC atrophy. This paper argues that HPC atrophy, if it does occur following vestibular loss, may not be directly, causally responsible for the cognitive deficits, and that it is more likely that rapid functional changes in the HPC are responsible, due to the interruption of the transmission of vestibular information to the HPC. The argument presented here rests on 3 tranches of evidence: (1) Cognitive deficits have been observed in humans even in the absence of HPC atrophy; (2) HPC atrophy has not been reported in animal studies following vestibular loss, despite cognitive deficits; and (3) Animal studies have shown that the interruption of the transmission of vestibular information to the HPC has immediate consequences for HPC place cells, far too quickly to be explained by HPC atrophy. It is possible that HPC atrophy, when it does occur, is related to the longer-term consquences of living with vestibular loss, which are likely to increase circulating cortisol.

show abstract

Secreted Glycoproteins That Regulate Synaptic Function: the Dispatchers in the Central Nervous System

Gong,

Zhu,

Han

et al. 2023

Mol Neurobiol

View full text Add to dashboard Cite

Rethinking the hippocampal cognitive map as a meta-learning computational module

Cited by 8 publications

References 66 publications

In Search of Dispersed Memories: Generative Diffusion Models Are Associative Memory Networks

In Search of Dispersed Memories: Generative Diffusion Models Are Associative Memory Networks

Interpreting the meaning of changes in hippocampal volume associated with vestibular loss

Secreted Glycoproteins That Regulate Synaptic Function: the Dispatchers in the Central Nervous System

Contact Info

Product

Resources

About